| """Custom CUDA/Triton kernels for SEM V6 optimization.""" | |
| from .fused_mhc_kernel import ( | |
| fused_width_connection_triton, | |
| compare_with_einops_reference, | |
| benchmark_kernel_speedup, | |
| ) | |
| __all__ = [ | |
| "fused_width_connection_triton", | |
| "compare_with_einops_reference", | |
| "benchmark_kernel_speedup", | |
| ] | |